Researcher Workflows
Phân tích code, xử lý documents, research kỹ thuật.
31. Phân tích codebase lạ
“Cần hiểu repo open-source 50k+ lines để contribute”
Bước 1 — Generate overview
/mk:docs init
Kit scan toàn bộ repo: file structure, dependencies, entry points, patterns → tạo docs.
Bước 2 — Deep dive
Hỏi cụ thể:
Giải thích kiến trúc plugin system trong repo này
Vẽ diagram data flow từ API request đến database
/mk:preview --diagram "request lifecycle from HTTP to database"
Bước 3 — Tìm entry point để contribute
/mk:plan "add support for WebSocket connections to the existing HTTP server"
Kit phân tích existing architecture, đề xuất cách thêm feature mà không break patterns.
Tips
- Bắt đầu bằng
docs/system-architecture.mdsau khi chạy/mk:docs init. - Hỏi “file nào handle X?” trước khi đọc code — kit trả lời nhanh hơn đọc thủ công.
32. Xử lý documents hàng loạt
“100 PDF invoices, cần extract data thành spreadsheet”
Bước 1 — Extract từ PDF
/mk:cook "read all PDF files in ./invoices/, extract: invoice number, date, amount, vendor name. Save as invoices.xlsx"
Kit dùng pdf skill để: parse PDF → extract text/tables → structure data → export Excel.
Bước 2 — Verify
Kiểm tra 5 dòng đầu trong invoices.xlsx có đúng với PDF gốc không
Các format khác
Word → structured data:
/mk:cook "extract all section headings and tables from contract.docx, save as markdown"
Excel analysis:
/mk:cook "analyze sales.xlsx: monthly revenue trend, top 10 products, year-over-year comparison. Create summary report."
PowerPoint → docs:
/mk:cook "extract all text and speaker notes from presentation.pptx, save as markdown document"
Lưu ý
Cần cài Python dependencies trước:
cd .claude/skills && bash install.sh
33. Research kỹ thuật
“So sánh 3 database options cho project mới”
Bước 1 — Research
/mk:brainstorm "database for real-time analytics: PostgreSQL vs ClickHouse vs TimescaleDB. Consider: query speed, write throughput, operational complexity, cost, ecosystem."
Kit phân tích pros/cons từ nhiều góc: performance, operations, cost, developer experience.
Bước 2 — Structured analysis
/mk:plan "evaluate and select database for real-time analytics dashboard"
Kit tạo comparison matrix:
| Criteria | PostgreSQL | ClickHouse | TimescaleDB |
|---|---|---|---|
| Write throughput | ★★★ | ★★★★★ | ★★★★ |
| Query speed | ★★★ | ★★★★★ | ★★★★ |
| Ops complexity | ★★ | ★★★★ | ★★★ |
| Cost | ★★ | ★★★ | ★★★ |
| Ecosystem | ★★★★★ | ★★★ | ★★★★ |
Bước 3 — Risk assessment
/mk:predict "choosing ClickHouse for real-time analytics in a small team"
5 experts flag risks: operational overhead, hiring difficulty, migration complexity.
Bước 4 — Document decision
/mk:docs update
Kit lưu decision record vào docs/ — lý do chọn, alternatives considered, trade-offs accepted.